Computer implemented method, processor device and computer program product for designing intervention into real complex systems

1. A computer-implemented method for designing intervention into the behavior of a real complex system of biochemical nature, the method comprising: a) modeling the real complex system by a network of objects and relations between said objects, wherein said objects of the system are represented by network points and relations are represented by edges between the network points, and wherein the states of the objects are described by a parameter set and the relations associated with the edges are described by functions of time; b) for each object of the real system, obtaining values for each parameter of said parameter set both for an initial state and a desired target state thereof; c) setting the initial values and the desired target values of the parameters of the network points; d) using a predetermined metaheuristic algorithm, automatically generating an initial set of test excitations for at least one point of the network, the initial set of test excitations including a predetermined number of test excitations defined by the metaheuristic algorithm; e) simulating the behavior of the network by using the set of test excitations; f) detecting whether a simulation termination condition in a given simulation step is true and if so, stopping the simulation; g) after stopping the simulation, calculating and storing, for each network point, the difference between the parameter values belonging to the desired target state and the parameter values produced by the simulation; h) automatically generating a next set of test excitations by using the predetermined metaheuristic algorithm, the next set of test excitations being generated to more closely approach the target state based on the previous test excitations having the smallest difference between the parameter values belonging to the desired target state and the parameter values produced by the simulation, and the number of text excitations in the next set being defined by the metaheuristic algorithm; i) iteratively repeating steps e)-h) to reduce the difference between the parameter values produced by the simulations and the parameter values of the target state until a predetermined entire termination condition is satisfied; j) from among the stored simulation results, selecting the simulation result best matching the desired target state, the test set of excitations which produced said best matching simulation result is regarded as a final excitation set transferring the network from its initial state to its target state; and k) outputting the final excitation set as the intervention into the behavior of the real complex system of biochemical nature.

2. The method according to claim 1 , wherein the initial values and the desired target values of the parameters of the objects are determined by measurements on the objects.

3. The method according to claim 1 , wherein the initial values and the desired target values of the parameters of the objects are read from a database.

4. The method according to claim 1 , wherein a simulation time window is defined in advance and the simulation termination condition is regarded satisfied if within a sufficiently long time window, a steady state of the network is detected.

5. The method according to claim 4 , wherein the resulting steady states are stored, and wherein the method further comprises outputting a list of the stored steady states and the number of simulations reaching each steady state.

6. The method according to claim 5 , wherein all of the states within a specific tolerance range are regarded as the same state.

7. The method according to claim 1 , wherein the predetermined entire termination condition comprises obtaining a difference value between the parameter values belonging to the desired target state and the parameter values produced by the simulation that is below a predetermined threshold.

8. The method according to claim 1 , wherein the predetermined entire termination condition comprises reaching a predetermined number of simulation iterations.

9. The method according to claim 1 , wherein the metaheuristic algorithm is a genetic algorithm, the number of test excitations is more than one per simulation, a fitness value for each simulated set of test excitations is calculated by the difference between the parameter values belonging to the desired target state and the parameter values produced by the simulation, and the next set of test excitations is generated from the sets of test excitations having the best fitness values by crossing over and/or mutation.

10. The method according to claim 1 , wherein the metaheuristic algorithm is a simulated annealing algorithm, the number of test excitations is one per simulation, and an energy value for each simulation iteration is calculated by the difference between the parameter values belonging to the desired target state and the parameter values produced by the simulation.

11. The method according to claim 1 , wherein the intervention is a targeted drug, the real complex system of biochemical nature is a cell, the initial state is a phenotype belonging to a particular disease, the target state is a healthy state, and the final excitation set comprises the shortest chain of actions between the initial state and the target state.

12. A processor device for designing intervention into the behavior of a real complex system of biochemical nature, the device comprising: a data carrier storing a model of a real complex system of a biochemical nature, the model comprising a network of objects represented by network points and relations between the objects being represented by edges between the network points, wherein the states of the objects are described by a parameter set and the relations associated with the edges are described by functions of time; a processor device; and a memory device encoding instructions that, when run on the processor device, cause the processor device to: (a) obtain values for an initial state and a desired target state for each parameter of the parameter set for each object; (b) set the initial values and the desired target values of the parameters of the network points; (c) automatically generate an initial set of test excitations for at least one point of the network, the initial set of test excitations having a number of test excitations selected using a predetermined metaheuristic algorithm; (d) simulate behavior of the network using the set of test excitations; (e) stop the simulation when a simulation termination condition in a given simulation step is detected; (f) after the simulation is stopped, calculate and store the differences between the parameter values belonging to the desired target state and the parameter values produced by the simulation for each network point; (g) automatically generate a next set of test excitations selected to more closely approach the target state based on the previous test excitations having the smallest differences by using the predetermined metaheuristic algorithm; (h) iteratively repeating steps (d)-(g) until a predetermined termination condition is satisfied; (i) from among the stored simulation results, selecting the simulation result best matching the desired target state, the test set of excitations which produced said best matching simulation result is regarded as a final excitation set transferring the network from its initial state to its target state; and (j) outputting the final excitation set as the intervention into the behavior of the real complex system of biochemical nature.

13. The processor device of claim 12 , wherein the intervention is a targeted drug, the real complex system of biochemical nature is a cell, the initial state is a phenotype belonging to a particular disease, and the target state is a healthy state.

14. A non-transitory computer program product comprising computer readable instructions which, when run on a computer, cause the computer to: a) model a biochemical system with a network of objects represented by networks points and relations between the objects represented by edges between the network points, wherein the states of the objects are described by a parameter set and the relations associated with the edges are described by functions of time; b) for each object of the biochemical system, obtain values for each parameter of the parameter set for an initial state and a desired target state thereof; c) set the initial values and the desired target values of the parameters of the network points; d) automatically generate an initial set of test excitations for at least one point of the network using a predetermined metaheuristic algorithm, the initial set of test excitations including a predetermined number of test excitations defined by the metaheuristic algorithm; e) simulating the behavior of the network using the set of test excitations; f) detect whether a simulation termination condition in a given stimulation step is true and if so, stop the simulation; g) after the simulation is stopped, calculate and store, for each network point, the difference between the parameter values belonging to the desired target state and the parameter values produced by the simulation; h) based on the differences and number of test excitations, automatically generate a next set of test excitations using the predetermined metaheuristic algorithm, the next set of test excitations being generated to more closely approach the target state based on the previous test excitations having the smallest difference between the parameter values belonging to the desired target state and the parameter values produced by the simulation, and the number of text excitations in the next set being defined by the metaheuristic algorithm; i) iteratively repeat steps e)-h) to reduce the difference between the parameter values produced by the simulations and the parameter values of the target state until a predetermined termination condition is satisfied; j) from among the stored simulation results, select the simulation result best matching the desired target state, where the test set of excitations which produced the best matching simulation results is regarded as a final excitation set transferring the network from its initial state to its target state; and k) outputting the final excitation set as a designed intervention into the biochemical system.

15. The computer program product of claim 14 , wherein the biochemical system is a cell, the initial state corresponds to a phenotype of a given cell type, and the target state is either a healthy or an apoptotic state.

16. The method according to claim 14 , wherein the metaheuristic algorithm is a genetic algorithm and a fitness value for each simulation iteration is calculated by the difference between the parameter values belonging to the desired target state and the parameter values produced by the simulation.

17. The method according to claim 14 , wherein the metaheuristic algorithm is a simulated annealing algorithm and an energy value for each simulation iteration is calculated by the difference between the parameter values belonging to the desired target state and the parameter values produced by the simulation.